2012-2013 Bowl Pick'em Picks: Aggressive / Midsize Pools

Origin

Our picks are computer-optimized picks. To create them, we built a computer simulation engine that plays out thousands of hypothetical bowl pick’em contests. For each of these contests, we first simulate how a panel of fictional pool opponents will make their picks, then we project how the 2012-2013 bowl season will play out.

The last step is to iterate through a vast array of potential pick sets you could play in these hypothetical pools, in order to find out which pick sets most often outperform the fictional opponents. This pick set was one of the winners for midsize pools.

The entire process takes days to run and incorporates a variety of objective data, including public picking trends (published by several large sports sites, but adjusted by us for some artificial biases), Vegas odds, and game predictions from our proprietary mathematical models.

The Data Rules

A lot of people have preconceptions about how to approach bowl pick’em contests. Those preconceptions are often based on what’s happened in past bowl seasons, what the popular sports media are saying about this year’s games, or gut instinct about “intangibles” such as which teams are more likely be motivated for their bowl games.

The objective approach we apply, on the other hand, has no preconceptions. Instead, it evaluates hard data to uncover whatever inefficiencies happen to exist in each unique bowl season. Every year, the dynamics of bowl contests are different. There can be lots of evenly matched teams, or lots of lopsided matchups; lots of big name teams playing no-name schools, or few such games. Oh, and we always love to see games featuring mediocre teams from big name conferences. These types of factors will impact your optimal bowl pick’em strategy.

If you blindly apply the same tactics year after year (e.g. “With 35 games, I’m going to pick at least four upsets”), you’re putting yourself at a huge disadvantage. If your goal is to win your bowl pool, all that matters is making decisions that increase your chances of winning. And with billions upon billions of potential combinations of bowl picks to make, computer simulation is an ideal solution to identify pick sets that should give you a big edge.

So as you review these picks, feel free to ask us questions about why we did this or didn’t do that. After researching contest strategy for many years now, we can usually explain why a particular idea makes sense or not. However, for more nuanced questions, sometimes the answer is simply, “Because we simulated a similar approach for pools of this size, and it didn’t perform nearly as well, no matter how nice it sounds in theory.”

General Strategy For Midsize Pools

Once we get above 50 people into the 100+ person range, coming in top 10% probably isn’t going to cut it anymore, unless you’re in a pool that offers prizes to more than the top five finishers or so. So maximizing our chances to finish in the money involves taking more risk. If a couple more key games go our way, we should do very well; if they don’t, we’ll likely end up with a worse finish than the Small Pool strategies.

In some years, graduating from a Small Pool to a Midsize Pool strategy involves picking at least several more upsets, typically of the lower-risk variety. However, that’s not really possible this year (see the following section). So an alternative way to increase the aggressiveness of a pick set is to take several smaller risks by shifting teams up and down the confidence point scale.

A Quick Thought About Upsets

Still, we tend to get a decent amount of feedback related to upsets, or more commonly, about the perceived lack of upsets in our various strategies.

One reality is that with 35 winners to pick and confidence points to assign to each, the majority of your competitors have way too much information on their hands to process anyhow. They’re going to make a bunch of suboptimal decisions in terms of assigning confidence points, and they’re also likely to get riskier than they should regarding picking upsets themselves. So we prefer to stand by and watch as they shoot themselves in the foot.

Another reality, specific to this year, is that the 2012-13 college bowl season is projected to be one of the most lopsided in recent memory. As a result of many games having a solid favorite, the list of underdogs with a great shot to pull off an upset is thinner than usual. So our strategies this year had to recognize and adjust to this case; it’s much harder to find high-value, low-risk upset picks in this year’s bowls.

Confidence Points Make Things Trickier

Evaluating value pick opportunities in confidence point based pools is a lot more complex than for non-confidence pools. In a traditional pool, it’s pretty clear how to act on value scenarios. If BYU has 55% win odds against San Diego State, but only 45% of your opponents are picking BYU to win, with few exceptions, BYU is going to be a good pick in a non-confidence pool.

Confidence points adds a new dimension, though. What if the 45% of people picking BYU to win have all done so with the maximum of 35 confidence points, and the 55% picking SDSU have also done so with 35 points? In this case, taking BYU with much lower confidence than 35 may be the best call, but it also depends on what else you’ve got going on with other picks.

Notes About This Pick Set

• Every team whose win odds are at least 15% lower than their public pick rate is at least 6 confidence points lower (and all the way up to 16 points lower in the case of LSU, Stanford, and San Jose State) than they would be in a pure ranking of our projected win odds. That includes a batch of five teams that all have projected win odds between 65% and 69% and pick rates between 84% and 88% (includes LSU, West Virginia, Cincinnati), plus a couple large favorites (Florida and Georgia).
• The most aggressive pick in this set may be Arkansas State at 34 confidence points. Arkansas State has only middle-of-the-pack win odds (68%), and are just four point favorites. But the public is tilted so heavily towards Kent State (71% pick rate), that this game presents a great chance to gain ground on a lot of opponents.

A Closing Point On Expectations

An average picker will win a 75-person pool once every 75 years -- in other words, maybe once during their bowl pick’em playing career. Our ROI projections indicate these picks should increase that win frequency to once every 20-25 years or so.